Software Engineer, Tokens and Prompt Structures

Job Overview

As a Software Engineer on the Encodings Infra team at Anthropic, you will own the design and maintenance of libraries that engineers and researchers use to encode text and multimodal data for consumption by Claude, our AI model. Your work will have a broad impact across the codebase, collaborating with researchers and engineers to advance our encoding ideas from experiment to production.

Responsibilities

• Maintain and improve encoding libraries used across Anthropic.
• Run experiments to determine optimal ways to feed structured data into Claude.
• Design data structures and abstractions that simplify encoding complexities for most of the organization while empowering power users.
• Adapt encoding libraries to support new research directions and ensure their production readiness.
• Optimize encoding performance across dependent systems.

Requirements

• 5+ years of software engineering experience, particularly with library, SDK, or developer-facing API maintenance.
• Familiarity with ML terminology and LLM architecture.
• Experience with complex refactors in large codebases.
• Strong communication skills for collaboration with researchers and engineers.
• Results-oriented mindset with flexibility and impact.
• Care about the societal impacts of your work.

Nice to Have

• Experience with tokenizers or other text/data encoding systems.
• Long-term maintenance of widely-used libraries.
• Performance optimization expertise.
• Proficiency in Python and/or Rust.
• Background in reinforcement learning or model training infrastructure.

Compensation

The annual salary range for this role is $320,000-$405,000 USD.